Voice feature extraction from agitated speech

This research paper aims to identify voice features by literature survey and validate them for human agitation detection. The prosodic features pitch and loudness were identified through literature survey and then voice snippets were analysed and tested in offline (pre-recorded) speech for validation. These voice features are correlated with quantifiable sound wave metrics such as frequency and amplitude, respectively. By identifying distinct variations in these metrics, agitation behaviour can be extrapolated. Using the results found, the agitation state of the person is to be analysed. Such an agitation detection system can benefit in improving the quality of life of caregivers and elderlies.